A general ranking strategy for data accuracy management

A series of recent studies proposed a construct named damage and a set of models for estimating the damage in a chosen class of information systems. The perception that underlies the proposed construct is that, all other things being equal, it would be beneficial to assign priority to the elimination of data errors that have a stronger negative effect on output accuracy (i.e., output accuracy is lower) over data errors that have a weaker effect. In this paper we extend the work on damage by considering its use with information systems in general, rather than a specific class of information systems. Mainly, we propose a general strategy for ranking the inputs according to the damage that errors in each input inflict on the output of the system. A major advantage of this strategy is that it focuses the ranking effort on a subcomponent of the information system that can be substantially smaller and simpler than the information system as a whole.